Our study is the first evaluation of the relationship between residential concentrations of PCBs and risk of ALL in children. We observed increased risk of ALL with increasing concentrations of total PCBs and with specific PCB congeners in dust samples taken from the room in which the child spent the most time. In contrast, dust levels of the persistent organochlorine pesticides DDT, DDE, chlordane, methoxychlor, and pentachlorophenol were not associated with increased risk. PCBs are probable human carcinogens and cause perturbations of the immune system, and concentrations in house dust and plasma have been associated with increased risk of adult NHL in epidemiologic studies (Colt et al. 2005
; De Roos et al. 2005
; Engel et al. 2007
). Our findings suggest that residential exposure to PCBs may represent a previously unrecognized risk factor for the development of ALL in young children.
The associations we observed with specific PCB congeners were consistently stronger among non-Hispanic whites compared with Hispanics. We were unable to explain this effect modification, although unmeasured factors such as cultural or biological characteristics may have been responsible. It is also possible that residential dust samples were not an important source of exposure among Hispanic children in our study or that other unmeasured sources of exposure were more important, resulting in misclassification that would have obscured positive associations with risk. For example, Hispanic children may have spent more time away from their home in the care of relatives. We did not have information on the usual amount of time spent in the home or in the room where we conducted the dust sampling, which would have partly addressed this question.
Residential exposure to PCBs and childhood leukemia risk has not been evaluated previously in a population-based study. A small study comparing bone marrow concentrations of PCBs in 38 children with ALL and 15 healthy children selected from the same hospital found no significant differences in mean concentrations across the two sample pools (Scheele et al. 1992
). That study was limited by a small size, measurement of PCBs in samples collected postdiagnosis, and use of a comparison group that may not have been representative of the general population. Case–control and cohort studies of adult NHL in the general population have shown consistent positive associations between PCB concentrations and risk (Engel et al. 2007
). A population-based case–control study measuring PCBs in both house dust and plasma samples found increased NHL risk for both exposure assessment methods (Colt et al. 2005
; De Roos et al. 2005
), suggesting that house dust may be a good exposure indicator even in an older population.
We found no associations with childhood ALL for the organochlorine pesticides. Although self-reported parental occupational and residential exposures to pesticides have been associated with increased ALL risk in many studies (Buffler et al. 2005
; Daniels et al. 1997
; Infante-Rivard and Weichenthal 2007
; Zahm and Ward 1998
), nearly all of them lacked information on specific pesticide active ingredients. A study in Costa Rica (Monge et al. 2007
) reported a nonsignificantly elevated risk of childhood leukemia associated with organochlorine pesticide use by the father during the pregnancy; however, results were based on small numbers, and ORs for specific pesticides were not presented. Among a cohort of agricultural pesticide applicators, paternal preconception use of aldrin was associated with an elevated risk of all childhood cancers; however, the numbers were too small to evaluate specific cancer types (Flower et al. 2004
). Occupational exposure to chlordane has been linked to an elevated risk of adult leukemia in some studies (Brown et al. 1990
; Purdue et al. 2007
). In contrast to our findings, two case series reported blood dycrasias and leukemias after professional pest control treatments of residences with chlordane (Epstein and Ozonoff 1987
; Infante et al. 1978
), providing limited evidence that residential chlordane exposure may disrupt the immune system in humans.
Children may be exposed to PCBs and persistent organochlorine pesticides in utero
, through breast-feeding and other dietary sources, through inhalation, and through ingestion of house dust (Calabrese et al. 1989
; Guvenius et al. 2003
; Hooper et al. 2007
; Wilson et al. 2001
). Young children are estimated to ingest about 100 mg of dust per day (Davis et al. 1990
). Levels of PCBs and other organochlorine chemicals in blood, breast milk, and dietary sources have decreased considerably since the early 1970s (Furst 2006
; Laden et al. 1999
; Schecter et al. 2005
). Accordingly, exposure to environmental sources such as dust and soil via ingestion, inhalation, and dermal absorption might account for an increasingly large share of children’s exposure to these chemicals. A study of preschool children in the United States found that indoor air and dietary sources were responsible for most exposures to PCBs (Wilson et al. 2001
). PCB concentrations in carpet dust may be a good predictor of indoor air concentrations because carpets act as a reservoir for continued volatilization, and chemicals absorbed to suspended particles can be inhaled (Butte and Heinzow 2002
). Sources of PCBs in carpet dust, particularly in older homes, include paints, sealants, caulking, floor finishing products, and older light fixtures (Herrick et al. 2004, 2007
; Rudel et al. 2008
Although PCB concentrations in carpet dust were associated with increased leukemia risk, PCB loadings were not. The loading incorporates the concentration and amount of dust collected per area of carpet sampled (i.e., concentration multiplied by the total dust collected per square meter of carpet) and is postulated to be a more accurate indicator of exposure for small children (Bradman et al. 1997
; Lanphear et al. 1996
). However, the amount of dust collected is likely to reflect recent vacuuming practices, and a single measurement of dust loading would not be reliable if there is a large variation in dust loading within a house or over time. Thus, incorporating the dust loading into the exposure metric may have introduced random error and obscured concentration-based associations with risk.
Our exposure assessment method is a major strength of our study. We conducted environmental sampling to measure concentrations of persistent organochlorine chemicals in the room in which the child spent the most time. Although it is desirable to do so, prospective or pretreatment exposure measurements in blood samples are difficult to obtain because of the rarity of the disease and the very young age of most children. Residential carpet dust sampling provides an alternative exposure assessment method that allows identification of individual compounds and is not subject to recall bias. Organochlorine pesticides and PCBs are known to persist in carpets, where they are protected from degradation by sunlight, moisture, and microbes, and most homes had detectable levels of these chemicals in the dust samples. Other strengths of our study include the rapid case ascertainment, population-based selection of controls, and high participation rates for the tier 2 interview.
A limitation of our study was the moderate number of participants with HVS3 dust samples, particularly after stratifying by ethnicity, limiting our statistical power. Further, our study population was limited to those who had not moved since the diagnosis or reference date. If PCB and organochlorine concentrations were substantially different among those who were less residentially stable, our results may not be generalizable to the general population. Another limitation, common to many case–control studies, was the lower response rates among controls compared with cases. Our population-based controls were previously shown to be comparable with all eligible controls with respect to maternal history of fetal loss, birth weight, birth order, and time since last live birth (Ma et al. 2004
). In the present analysis, cases and controls in all ethnic groups were significantly different with respect to household income. However, income was not associated with PCB concentrations and was not a confounder in our analyses. Moreover, analyses stratified by income showed similar associations between PCB concentrations and ALL risk, providing some assurance that our findings were not caused by selection bias. The fact that we did not observe associations for other persistent organochlorine chemicals provides additional evidence against selection bias.
The positive association observed for PCBs and ALL could have been attributed to other chemicals in the home that were highly correlated with PCBs. Although we were able to rule out confounding by the organochlorine pesticides measured in this study, we did not measure concentrations of polybrominated diphenyl ethers (PBDEs), which have become common environmental contaminants because of their increased use as fire retardants since the 1970s. Several studies (Bradman et al. 1997
; Furst 2006
; She et al. 2007
), including a cohort study of Hispanic women with young children in Salinas County, California (Bradman 2007
), found no correlation between PCB-153 and PBDE concentrations in maternal blood samples, suggesting different routes of exposure for these classes of chemicals. However, future analyses of residential concentrations of organochlorine chemicals, PBDEs, and other pesticides will be informative.
In summary, we observed an increasing risk of ALL associated with increasing residential concentrations of PCBs. PCBs are considered probable human carcinogens and cause perturbations of the immune system. We found no evidence of a relationship between ALL risk and exposure to DDT, DDE, chlordane, or pentachlorophenol. Additional studies are needed to further evaluate these highly suggestive findings.